Estimation of a constrained multinomial logit model

摘要: Identifying the set of available alternatives in a choice process after considering an individual’s bounds or thresholds is complex that, practice, commonly simplified by assuming exogenous rules formation. The Constrained Multinomial Logit (CMNL) model incorporates several attributes as key endogenous to define choice/rejection mechanism. allows for inclusion multiple constraints and has closed form. In this paper, we study estimation CMNL using maximum likelihood function, develop methodology estimate overcoming identification problems partition sample, test with both synthetic real data. appears be suitable general applications it presents significantly better fit than MNL under constrained behaviour replicates estimates unconstrained case. Using mode data, found significant differences values times elasticities between compensatory semi-compensatory models, which increase on become active.